JP2009240569A - Rib area detecting apparatus and method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To detect each of the plurality of divided rib areas by separating respective rib areas connected by calcification or the like. <P>SOLUTION: In the method of detecting the rib area by labeling each rib area of a subject from a three-dimensional image indicating the subject, whether or not it is an area detected as one rib area by the connection of the plurality of rib areas is decided on the basis of the shape information of each detected rib area. When it is decided that it is the area detected as one rib area by the connection of the plurality of rib areas, the detected rib area is divided into the plurality of rib areas configuring one rib area, and when the division into the plurality of rib areas is made, by labeling the plurality of divided rib areas respectively, the rib area is detected again. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a rib detection process, and more particularly to a rib area detection apparatus and method suitable for determining whether or not a rib erroneously detected area exists in a three-dimensional image, and a program therefor.

  Conventionally, in the medical field, each rib area is automatically detected based on a three-dimensional image composed of a plurality of medical images scanned from a chest to the base of a foot at a plurality of cutting positions. This is an indispensable technique for observing a region and other anatomical feature regions existing around it.

Therefore, a process of automatically detecting using a computer is performed by labeling each rib region from a three-dimensional image. For example, Patent Document 1 proposes a technique for automatically detecting a rib region from a three-dimensional image composed of CT volume data and labeling the rib from the size, shape and relative positional relationship of the detected rib. ing.
JP 2007-111534 A

  However, in the above method, the ribs of a normal subject as shown in FIG. 2A are calcified, and the subject's ribs and ribs are connected as shown in R1 in FIG. 2B, and the same applies to different ribs. There is a problem in that the rib area cannot be accurately detected due to the labeling number.

  Therefore, in view of the above circumstances, the present invention has an object to provide a rib area detection device and method that enables redetection of a rib area based on the detected shape information of each rib area, and a program therefor. To do.

  The rib area detecting device of the present invention includes rib area detecting means for detecting a rib area by labeling each rib area of a subject from a three-dimensional image representing the subject. Based on the shape information of the rib area, a determination means for determining whether or not the multiple rib areas are areas detected as one rib area, and the multiple rib areas are connected by the determination means And a dividing unit that divides the rib region into a plurality of rib regions that constitute one rib region when it is determined that the rib region is detected as one rib region. When division into a plurality of rib areas is performed, the rib areas are re-detected by labeling each of the divided rib areas. .

  The rib area detection device of the present invention includes a statistical database having statistical height height information in the body axis direction, and the detected height width of each rib area in the body axis direction A comparison means for comparing whether or not the height width information of the rib is greater than a predetermined value, and when the determination means determines that the height is greater than the predetermined value by the comparison means, It may be determined that the rib areas are detected as one rib area by connecting the rib areas.

  Furthermore, “the height width information of the ribs in the body axis direction of the statistical database” may be a numerical value set as the maximum width of the ribs in the body axis direction on the statistics.

  The rib area detecting device of the present invention includes a calculating means for calculating an average width of each detected rib area from the height width in the body axis direction of each detected rib area, and a body axis of each detected rib area. Comparing means for comparing whether or not the height width in the direction is larger than the calculated average width of the rib area by a predetermined value or more, and the determining means is the rib area calculated by the comparing means When it is determined that the average width is greater than a predetermined value, a plurality of rib regions may be determined to be a region detected as one rib region by connecting them.

  Further, the “predetermined value” may be a value obtained by subtracting the average width calculated from the width estimated as the maximum height width from the detected height width of each rib region in the body axis direction.

  The rib area detection device of the present invention includes a statistical database having statistical rib shape information, and the detected shape of each rib area has an allowable range as a rib shape from the statistical rib shape information of the statistical database. A comparison means for comparing whether or not the rib area exceeds the plurality of ribs when the determination means determines that the shape of the rib area detected by the comparison means exceeds the allowable range as the rib shape. It may be determined that the region is a region detected as one rib region by connecting the regions.

  Furthermore, the “statistical rib shape information” may be any one of the perimeter of the ribs on the cross-sectional image of the statistical rib, the statistical rib boundary complexity, and the statistical rib thickness. Or one of them.

  The “three-dimensional image representing the subject” is data representing the internal structure of the subject by the values of a plurality of pixels arranged three-dimensionally.

  The rib region detection method of the present invention is a method for detecting a rib region by labeling each rib region of a subject from a three-dimensional image representing the subject, and a plurality of ribs based on the detected shape information of each rib region. It is determined whether or not the region is a region detected as one rib region by connecting, when it is determined that the region is detected as one rib region by connecting a plurality of rib regions, When the rib area is divided into a plurality of rib areas constituting one rib area and divided into a plurality of rib areas, the rib area is re-detected by labeling each of the divided rib areas. Features.

  The program for detecting a rib area of the present invention is based on the detected shape information of each rib area in a program for detecting a rib area by labeling each rib area of a subject from a three-dimensional image representing the subject on a computer. A function of determining whether or not a plurality of rib areas are areas detected as one rib area, and an area detected as one rib area by connecting a plurality of rib areas. When it is judged that there is a function to divide into a plurality of rib areas constituting one rib area, and when divided into a plurality of rib areas, each of the divided rib areas is labeled. And a function of re-detecting the rib region.

  According to the rib area detecting device, method and program of the present invention, a plurality of rib areas are connected as a single rib area based on the shape information of each rib area detected by the rib area detecting means. It is determined whether or not it is a region, and when it is determined that the region is detected as one rib region by connecting the plurality of rib regions, the plurality of rib regions constituting the one rib region By dividing, it is possible to re-detect a region that is erroneously detected as one rib region by connecting a plurality of rib regions as a plurality of rib regions.

  Hereinafter, an embodiment in which a rib area detection device of the present invention is applied to one that extracts each rib area from a three-dimensional medical image will be described with reference to the drawings.

  The rib area detecting device shown in FIG. 1 detects the rib area by labeling each rib area of the subject from the image input means 1 for inputting the three-dimensional image acquired by the CT apparatus or the like and the three-dimensional image representing the subject. Whether or not it is a region detected as one rib region by connecting a plurality of rib regions based on the shape information of each rib region detected by the rib region detecting means 20 and the rib region detecting means 20 And a plurality of ribs constituting one rib area when the determination means 30 determines that the plurality of rib areas are detected as one rib area by connecting the rib areas. And dividing means 40 for dividing the area.

  The rib area detecting device includes a display unit 70 that can display a rib area detected from a three-dimensional image, and a database and a comparing unit to be described later.

  The configuration of the rib area detecting device is realized by executing a rib area detecting program read into the auxiliary storage device on a computer (for example, a personal computer).

  At this time, the target area extraction program is stored in an information storage medium such as a CD-ROM, or distributed via a network such as the Internet and installed in a computer.

  The schematic diagram of the rib in which the some rib area | region was connected by the calcification etc. which are shown under the image input means 10 shown in FIG. 1 is shown.

  As shown below the display means 70 shown in FIG. 1, the rib region detection apparatus divides a rib in which a plurality of rib regions are connected by calcification or the like by separating a portion connected by calcification or the like. It is possible to re-detect the rib area by labeling the plurality of rib areas.

  The rib area detecting means 20 detects the rib area by labeling each rib area of the subject from the three-dimensional image representing the subject. When the dividing means 40 divides the rib area into a plurality of rib areas, the rib area is detected. It is possible to re-detect the rib area by labeling each of the plurality of rib areas.

  For example, the rib region detection means 20 may use the technique of US Publication No. 2006-0062425 A1 (hereinafter referred to as Reference Document 1). Reference 1 discloses a high-speed tracing-based technique for extracting a rib region from a three-dimensional image. This technique is used to extract ribs and obtain separate data representing each rib. The detected ribs are ordered based on the size and shape of each rib.

  Further, the rib region detecting means 20 determines the order of the ribs by comparing the relative spatial relationship and the length ratio of the ribs. Typically, ribs do not have distinct features that distinguish each rib from each other, but ribs can be distinguished based on their relative size and shape. From the relative position of each rib seed point, the relative spatial order of the ribs from top to bottom and from left to right is evaluated. In addition, the symmetry axis of the rib cage is calculated from all the extracted centerline points of the ribs. Thereafter, the ribs are grouped in pairs in the direction of the symmetry axis according to the position of the ribs. After ordering and grouping, rib counting starts at the top of the rib cage and goes to the bottom. This is more complicated when the data has only part of the rib cage. The topmost rib is identified based on the fact that from the top to the bottom, the area surrounded by the rib pair expands and the shape of the area surrounded by the rib pair changes.

  Such a region and length ratio is used as a feature to identify the topmost rib. After the number of topmost ribs in the data set is identified, the remaining ribs are calculated and numbered from the topmost rib. Assign labels to the ordered ribs. Each rib is assigned a unique label that includes either the left or right side of the rib's body and the rib number of the rib (assigned from top to bottom on each side). For example, a label of “right rib No. 3” is assigned to the third rib from the top on the right side of the body.

  The rib area detecting means 20 displays the ribs on the display means 70 using such a label. When the user clicks on the rib using a mouse or the like (or simply keeps the pointer over the rib), the display means 70 can automatically display the rib number of the rib.

  Moreover, the rib area detection means 20 may be detected by labeling the rib area from the three-dimensional image by another method.

  For example, when detecting the rib area from the medical image, the rib area detecting means 20 generates an edge image from the medical image of the subject using an edge extraction filter and uses a Hough transform for detecting a parabola from the edge image. It is possible to use a method of detecting the rib shape by finding a parabola that seems to be a rib, or by detecting the rib region from the medical image by comparing the pixel value profile of the rib on the medical image with a predetermined threshold value. Good.

  Further, the rib area detecting means 20 may use a technique of detecting using a density histogram of the rib area, which will be described with reference to the drawings.

  FIG. 5 shows medical images of the vertebra and rib portions. First, the center line of the vertebra and the lateral width of the vertebra are obtained. Therefore, in the bone region in FIG. 5, the center line A in the longitudinal direction of the body surface (that is, perpendicular to the longitudinal direction of the body at the center of the longitudinal direction of the body in the cross section perpendicular to the direction of the body axis connecting the head and feet). A region below the horizontal line of the body is projected onto the horizontal axis to generate a density histogram.

  FIG. 6 shows a density histogram of the bone region generated in this way. In the histogram shown in FIG. 6, there is a large peak at the center and small peaks on the left and right. In the case of the bone region shown in FIG. 5, it is considered that the rib portion forms a small peak, the vertebra is where the most bone exists, and the largest peak is formed. Therefore, the maximum peak at the center is detected from the histogram of FIG. 6, and the vertebra region is detected as a corresponding portion.

  Then, as shown in FIG. 7, a circle S is applied to the upper part of the detected vertebra region 140 to determine the center C and the width d of the vertebra. Using the vertebra center C and the vertebra width d, the vertebra region, the left and right rib regions, and the sternum region are recognized.

As shown in FIG. 8, two half lines forming an angle of 60 ° upward from the vertebra center C obtained above are drawn, and the vertebra center C is included in the center so that the center line A is in the longitudinal direction of the body surface. Draw two perpendicular lines. The distance between the two straight lines is the width of the vertebra region, which is 1.5 of the width d.
To be doubled. In this way, the bone region of the medical image in FIG. 6 is divided into four regions by two half lines and two straight lines orthogonal to the center line A in the longitudinal direction of the body surface, and the vertebra region 140 and the right rib region 144 are respectively divided. , Sternum region 146 and left rib region 148.

  The rib area detecting means 20 recognizes the bone in units of parts such as the vertebra, the right rib, the sternum, and the left rib in this way. Then, different labels (numbers) are given to the recognized bone parts. The rib area detecting means 20 prepares a three-dimensional table for storing the recognition result, and gives the label to each voxel of the corresponding bone area. Further, the recognition result may be replaced with the three-dimensional data of the bone extraction result.

  In the above example, the bone region is recognized as a bone component such as the vertebra, the right rib, the sternum, and the left rib, but the right first rib, the right second rib, ..., the first cervical vertebra, The second cervical vertebra, the first thoracic vertebra, the second thoracic vertebra, and so on may be further divided.

  In this way, the bone area of each medical image is recognized in the same manner, and the bone shows a fairly high density value. Therefore, the gap where the density value is high is found and the ribs are recognized one by one in small parts. can do.

  In the display unit 70, a part corresponding to a part to be displayed is selected based on the part input by the user via the mouse or the keyboard.

  This selection by the user may be performed by selecting a part composed of a plurality of parts defined in advance.

  Note that the rib region detection means 20 may use a known method such as the above-described Patent Document 1 or the method of Japanese Patent Application No. 2006-280153 already filed.

  Hereinafter, the method for detecting the rib region of the present invention will be described with reference to the flowchart shown in FIG. First, the rib area detection unit 20 detects the rib area by labeling each rib area of the subject from the three-dimensional image representing the subject using the above-described method (step ST10).

  Thereafter, based on the shape information of each rib area detected by the rib area detecting means 10 by the judging means 30, it is determined whether or not the plurality of rib areas are connected to each other and detected as one rib area. Judgment is made (step ST20).

  FIG. 3 shows a panoramic projection image of the rib region viewed from the central axis of the subject.

  The horizontal axis is the angle viewed from the central axis of the subject, and the vertical axis is the body axis.

  The height distribution of the connected component indicated by the panorama projection image indicates a histogram of the height of the connected component in the rib region in the body axis direction and a statistic calculated from the histogram.

  Based on the comparison result of the comparison unit 30, the determination unit 30 determines whether or not the plurality of rib regions are areas detected as one rib region.

  The determination means 30 determines whether or not the plurality of rib areas are areas detected as one rib area by connecting them by any one of the first to third methods described below.

  First, the first method will be described.

  The rib area detecting device of the present invention includes a statistical database 60 having statistical height-width information in the body axis direction of the ribs, and the height width in the body axis direction of each detected rib area is the statistics of the statistics database 60. Comparing means 50 for comparing whether or not the height width information of the upper rib is greater than a predetermined value is provided.

  When the comparing unit 50 determines that the height of the ribs in the statistical database 60 is greater than a predetermined value, the area detected as one rib area by connecting a plurality of rib areas. It is notified to the judgment means 30.

  For example, the height width information of the rib in the body axis direction of the statistical database may be a numerical value set as the maximum width of the rib in the body axis direction on the statistics.

  The database 60 may calculate the maximum possible width of the rib from the height width information of the rib in the body axis direction by a computer, and estimate it as the calculated maximum width.

  For example, considering the resolution (slice spacing) of a medical image, to detect components that connect multiple rib regions that exceed the maximum width, the average width (μ), standard variance (σ), median Using (m), the following equations (1), (2), (3), etc. are taken into consideration for determination.

μ> MAX (1)
μ> MAX-σ (2)
m> MAX (3)
Next, the second method will be described.

  The rib area detecting device of the present invention includes a calculating means 80 for calculating an average width of each detected rib area from the height width in the body axis direction of each detected rib area, and a body of each detected rib area. Comparing means 50 for comparing whether or not the height width in the axial direction is larger than the calculated average width of the rib region by a predetermined value or more may be provided.

  In this case, when the comparison unit 50 determines that the average width of the rib region calculated by the calculation unit 80 is greater than a predetermined value, the comparison unit 50 is detected as one rib region by connecting a plurality of rib regions. The determination means 30 is notified that the area is the same.

  For example, the predetermined value may be a value obtained by subtracting the average width calculated from the width estimated as the maximum height width from the detected height width of each rib region in the body axis direction.

  For example, the average width (μi) and standard variance (σi) of the rib area calculated by the calculation means 80 are obtained, μm and σm are calculated for each calculated rib area, and μm + 2σm is estimated as the maximum height width. . When the average width is subtracted from the maximum height width, a predetermined value is obtained.

  Finally, the third method will be described.

  In addition, the rib area detecting device of the present invention includes a statistical database 60 having statistical rib shape information, and the detected shape of each rib area as a rib shape from the statistical rib shape information in the statistical database. Comparing means 50 for comparing whether or not the allowable range is exceeded may be provided.

  In this case, when the comparison unit 50 determines that the shape of the detected rib region exceeds the allowable range as the rib shape, the comparison unit 50 is a region detected as one rib region by connecting a plurality of rib regions. Is notified to the judging means 30.

  In addition, the statistical rib shape information is any one of the perimeter of the rib portion on the statistical cross-sectional image of the rib, the statistical rib boundary complexity, and the statistical rib thickness. Including one.

  For example, the complexity of the border of the ribs in the statistics is calculated using a known Fourier descriptor (for example, the Journal of Computer Aided Image Diagnosis Society Vol.5 No.2 Apr. 2001 <Chest Alternatively, it may be calculated using a quantitative analysis of lung cancer lesion candidate shadows in an X-ray CT image>.

  In addition, the thickness of the ribs in the statistics is a well-known secondary moment feature amount as a measure of thinness (for example, see the University of Tokyo Press September 10, 2004 <New Image Processing Handbook> pages 1534 to 1537) You may calculate using.

  Based on the notified comparison result, the determination unit 30 determines that a plurality of rib regions are connected to each other and are detected as one rib region.

Returning to the description of the flowchart shown in FIG. 4, when the dividing unit 40 determines that the plurality of rib regions are detected as one rib region by connecting the plurality of rib regions, the one rib region is selected. It divides | segments into the some rib area | region which comprises (step ST30).

  Various methods can be considered for the division. For example, one rib is sequentially extracted from the lowest rib area for a connected component in which a plurality of rib areas are connected.

  First, the bottom boundary line of the connected component is extracted. Since the upper and lower borders of the ribs are roughly parallel, it can be used to estimate the upper border of the ribs.

  In addition, since the upper and lower boundary lines of the space between the rib areas are substantially parallel, the boundary line on the lower side of the second lower rib can be obtained from the upper boundary line of the lower rib. . If the above process is repeated, all ribs can be separated.

  Finally, when the dividing means 40 divides the plurality of rib areas into the plurality of rib areas, the rib area detecting unit re-detects the rib areas by labeling the plurality of divided rib areas.

  As described above, according to the rib area detection device of the present invention, based on the detected shape information of each rib area, it is determined whether or not the plurality of rib areas are areas detected as one rib area. When a plurality of rib areas are determined to be an area detected as one rib area by connecting, the plurality of rib areas are divided into a plurality of rib areas constituting one rib area. As a result of the connection, a region erroneously detected as one rib region can be redetected as a plurality of rib regions.

The block diagram which shows one Embodiment of the rib area | region detection apparatus of this invention A diagram showing an example of a normal subject The figure which shows an example of the to-be-photographed object connected by each rib being calcified The figure which shows an example of the panoramic image of a rib area | region The flowchart which shows one Embodiment of the rib area detection method of this invention The figure which shows an example of the medical image of a rib area | region (the 1) Diagram showing density histogram of bone region below center line A of medical image The figure which shows an example of the medical image of a rib area | region (the 2) The figure which shows an example of the medical image of a rib area | region (the 3)

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Image input means 20 Rib area | region detection means 30 Judgment means 40 Separation means 50 Comparison means 60 Database 70 Display means

Claims (9)

In a rib area detection device comprising rib area detection means for detecting a rib area by labeling each rib area of the subject from a three-dimensional image representing the subject,
Based on the shape information of each rib area detected by the rib area detection means, a determination means for determining whether or not a plurality of rib areas are areas detected as one rib area;
Dividing means for dividing the plurality of rib areas into the plurality of rib areas constituting the one rib area when it is determined by the judging means that the plurality of rib areas are connected to each other and detected as one rib area. And comprising
The rib area detecting unit re-detects the rib area by labeling each of the plurality of divided rib areas when the rib unit is divided into a plurality of rib areas by the dividing unit. Detection device. A statistical database having statistical height-width information in the body axis direction of the ribs;
Comparison means for comparing whether or not the height width in the body axis direction of each of the detected rib regions is greater than a predetermined value more than the height width information of the statistical ribs in the statistical database, In addition,
The determination means determines that the plurality of rib areas are areas detected as one rib area by connecting the rib areas when the comparing means determines that the predetermined value is greater than a predetermined value. The rib area detecting device according to claim 1.   The rib area detecting device according to claim 2, wherein the height width information of the rib in the body axis direction of the statistical database is a numerical value set as the maximum width of the rib in the body axis direction on the statistics. A calculating means for calculating an average width of each detected rib area from a height width in the body axis direction of each detected rib area;
Comparing means for comparing whether or not the height width in the body axis direction of each of the detected rib areas is greater than a predetermined value more than the calculated average width of the rib areas,
When the determining means determines that the calculated average width of the rib area is greater than a predetermined value by the comparing means, the plurality of rib areas are connected and detected as one rib area. The rib area detecting device according to claim 1, wherein the rib area detecting apparatus determines that the area is an area.   The predetermined value is a value obtained by subtracting the calculated average width from a width estimated as a maximum height width from a height width in a body axis direction of each detected rib region. Item 5. A rib region detection device according to item 4. A statistical database with statistical rib shape information;
Comparing means for comparing whether or not the shape of each detected rib region exceeds the allowable range as the shape of the rib from the statistical rib shape information of the statistical database,
When the determining means determines that the shape of the detected rib area exceeds the allowable range as the rib shape by the comparing means, the plurality of rib areas are connected and detected as one rib area. The rib area detecting device according to claim 1, wherein the rib area detecting apparatus determines that the area is an area.   The statistical rib shape information may be any one of a perimeter length of a rib portion on a cross-sectional image of the statistical rib, a complexity of a boundary line of the statistical rib, and a thickness of the statistical rib. The rib area detecting device according to claim 6, wherein the rib area detecting device includes at least one of them. In a rib region detection method for detecting a rib region by labeling each rib region of the subject from a three-dimensional image representing the subject,
Based on the shape information of each detected rib area, it is determined whether or not it is an area detected as one rib area by connecting a plurality of rib areas,
When it is determined that the plurality of rib areas are areas detected as one rib area by connecting, the plurality of rib areas are divided into the plurality of rib areas constituting the one rib area,
A rib area detection method, wherein when the plurality of rib areas are divided, the rib areas are re-detected by labeling each of the divided rib areas. On the computer,
In a program for detecting rib areas by labeling each rib area of the subject from a three-dimensional image representing the object,
Based on the shape information of each detected rib area, a function of determining whether or not the plurality of rib areas are areas detected as one rib area,
A function of dividing the plurality of rib regions into a plurality of rib regions constituting the one rib region when determined to be a region detected as one rib region by connecting the plurality of rib regions;
A program for executing a function of re-detecting a rib area by labeling each of the plurality of rib areas divided when the plurality of rib areas are divided.